An Accurate Cancer Incidence in Barrett's Esophagus: A Best Estimate Using Published Data and Modeling

Abstract

Background & aims: Published estimates for the rate of progression from Barrett's esophagus (BE) to esophageal adenocarcinoma (EAC) vary. We used simulation modeling to reconcile published data and more accurately estimate the incidence of EAC among people with BE.

Methods: We calibrated the ERASMUS/UW model (a collaboration between Erasmus Medical Center, Rotterdam, the Netherlands and the University of Washington, Seattle) for EAC to match the 0.18% annual rate of progression from population-based studies. This model was then used to simulate the design of prospective studies, introducing more endoscopic surveillance. We used the model to predict rates of progression for both types of studies and for different periods of follow-up, and compared the predicted rates with published data.

Results: For the first 5 years of follow-up, the model reproduced the 0.19% mean annual rate of progression observed in population-based studies; the same disease model predicted a 0.36% annual rate of progression in studies with a prospective design (0.41% reported in published articles). After 20 years, these rates each increased to 0.63% to 0.65% annually, corresponding with a 9.1% to 9.5% cumulative cancer incidence. Between these periods, the difference between the progression rates of both study designs decreased from 91% to 5%.

Conclusions: In the first 5 years after diagnosis, the rate of progression from BE to EAC is likely to more closely approximate the lower estimates reported from population-based studies than the higher estimates reported from prospective studies in which EAC is detected by surveillance. Clinicians should use this information to explain to patients their short-term and long-term risks if no action is taken, and then discuss the risks and benefits of surveillance.

Keywords: Early Detection; Esophageal Cancer; Microsimulation; Population-Based Modeling.

Appendix Figure 1

Graphical representation of the natural history of EAC in the Erasmus/UW model.

Figure 1

Graphical representation of the Erasmus/UW Model

Figure 2

Diagnostic inaccuracy of endoscopic surveillance in the main analysis of our model

Figure 3

Cancer incidence rates for different lengths of follow up in the realistic scenario: A) average annual progression rates toward EAC and B) cumulative cancer incidence towards EAC

Figure 4

A) Cumulative cancer incidence rates per 1000 BE patients for 20 years of follow up in the population-based design without any surveillance (perfect surveillance scenario), in the population-based and prospective design for the realistic scenario. A) Cancer incidence rates per 1000 BE patients for 20 years of follow up in the population-based design without any surveillance (perfect surveillance scenario), in the population-based and prospective design for the realistic scenario.

Figure 5

High-grade dysplasia and cancer incidence rates for different lengths of follow up in the realistic scenario: A) average annual progression rates toward HGD and EAC and B) cumulative cancer incidence towards HGD and EAC

Figure 6

Cancer incidence rates for different lengths of follow up in the sensitivity analysis: Perfect surveillance scenario A) average annual progression rates toward EAC and B) cumulative cancer incidence towards EAC False-positive BE scenario C) average annual progression rates toward EAC and D) cumulative cancer incidence towards EAC Halting surveillance scenario E) average annual progression rates toward EAC and F) cumulative cancer incidence towards EAC